Bufetico Open Bufetico

What a P/E ratio actually says when memory prices hit a school's laptop budget

17 September 2026 · JH Grandgerard · 6 min read

The school superintendent in Upper Darby, Pennsylvania, is afraid his district can’t keep a laptop in every student’s hands much longer.

Dan McGarry jokes that before the pandemic, “a computer was a campfire” in his district: six kids around one machine, warming their hands, because that’s all the district could afford. Federal COVID relief money changed that. McGarry was out in line himself until 8:30 at night, handing out Chromebooks the district had rushed in overnight.

That money is gone. McGarry laid out the numbers for WHYY, Philadelphia’s public broadcaster, in a story in August. Chromebooks that used to cost about $400 a student now cost more than $600, and the district swaps out about 3,000 of them a year. He puts the cost of keeping up at $2 million to $3 million a year. Daniel Kitchen, the district’s technology director, said the average Chromebook goes up 25 percent a year, whoever sells it.

Going back to the campfire isn’t an option either. Pennsylvania gives its standardized tests online, and paper is only for students who need it as an accommodation.

The cause is memory. Benjamin Lee, who teaches computer science and electrical engineering at the University of Pennsylvania, explained it to the station. The big data center operators and AI companies are pouring hundreds of billions of dollars into data centers, and those have to be filled with hardware, memory included. Some of the plants that make memory chips, he said, can choose between chips for AI servers and chips for consumer electronics, and the more of the line goes to AI, the less room there may be for laptops. Lee said memory makers are getting profit margins he hasn’t seen in the 20 to 30 years he’s worked in the field, and that computer hardware is known for its booms and busts.

Micron, based in Boise, Idaho, is one of those memory makers, and its own numbers show those margins. In its quarter to the end of May, it kept 85 cents of every dollar of sales after covering the cost of making its products, against 38 cents a year earlier. Its profit for those three months, $28.2 billion, was more than three times the $8.54 billion it made in the whole of its 2025 fiscal year.

That profit is the E in P/E. A price-to-earnings ratio takes what the market says the whole company is worth and divides it by one year of profit: one number from the market, and twelve months of accounting. Reading a low P/E as a low price assumes those twelve months were a normal year. For a school district paying half again as much for every Chromebook, this year isn’t one.

Micron, year by year

Lee’s booms and busts are right there in Micron’s own filings. Every annual report gives a market value for Micron’s shares at the end of February or the start of March. Next to each one I put the profit for that whole fiscal year, which ends in late August or early September. So each ratio divides a market value from February or March by a year that was only half over at the time. You only get these exact ratios in hindsight. In billions of dollars:

Fiscal year Market value, Feb. or March Profit for the year Rough P/E
2018 45.0 14.14 3.2
2019 36.2 6.31 5.7
2020 44.8 2.69 16.7
2021 79.9 5.86 13.6
2022 83.9 8.69 9.7
2023 47.9 minus 5.83 none
2024 83.6 0.78 107
2025 85.7 8.54 10.0

In fiscal 2018 the ratio was 3.2. Profit fell 55 percent the next year and another 57 percent the year after, so the lowest P/E in the table sat right at the top of the profit cycle. In fiscal 2024, the thinnest year that still showed a profit, it was 107, the highest in the table, and the next year profit grew elevenfold. And fiscal 2023 has no P/E at all, because Micron lost money, so a screener sorting by P/E has no number for it in the year the cycle hit bottom.

The market is kind of guessing what the company makes in a normal year, and the price moves with that guess, slowly. The profit swings much harder, and Micron’s latest quarterly report even warns that in some past periods its prices fell below what its products cost to make. Divide a slow number by a fast one and the fast one runs the ratio. So on a cyclical company, one whose profit rides its industry’s booms and busts, a low P/E says this year’s profit is high and the market doesn’t believe it’ll last. A high one says the profit is low, and the market doesn’t believe that either. The ratio is reporting where the cycle is, upside down, and calling it a valuation. A P/E on Micron’s profit for the fiscal year that just closed would divide by the same months that are costing Upper Darby more than $600 a Chromebook.

One year against a whole stretch

Upper Darby buys a certain number of Chromebooks each year to replace the ones getting old or broken, Kitchen told WHYY. Buying that way also means no single year’s price decides what the whole fleet costs. When a company’s years aren’t alike, I put the price against its profit across a whole stretch, peaks and troughs together. That keeps the ratio from flipping every time the cycle turns, and that’s about all it does.

I still use the P/E. On a business whose profit moves a few percent a year, it’s a decent shortcut, because the ratio goes wrong in proportion to how much the E moves. So before you read one, look at the profit line for the last several years. If it’s a gentle slope, read the ratio. If it looks like a mountain range, the ratio is telling you which peak or valley you’re standing in.

Where this breaks

The value in each annual report leaves out the shares held by Micron’s executive officers and directors and by anyone holding five percent or more, so each ratio in the table is lower than a P/E on all the shares would be. Counting all of them, at the price on the day each year closed, could move every row. I stuck with what the company files, so you can redo the table from its annual reports.

Averaging across a cycle also needs a cycle to average. The table alone holds a year of losses and two years when profit fell by more than half, so Micron’s stretch means something. A business that hasn’t been tested yet, with one long boom and no bust, gives you a stretch that’s all peak. And one huge year can swamp the rest. The last two quarters Micron has reported, to the end of February and to the end of May, earned $13.8 billion and $28.2 billion, $42.0 billion together, more than the $41.2 billion that all the years in the table add up to. A stretch that takes in those two quarters leans on them too.

Dan McGarry and the other side of the memory boom

McGarry told WHYY he got into education partly to teach Shakespeare, and that the things he has to worry about now all cost money. Micron’s quarterly report splits its business into four units, and one of them makes memory and storage for phones and for PCs like the ones his district hands out. In the quarter to May, that unit earned $9.87 billion in operating profit, the profit from running it before interest and taxes. A year earlier it had earned $482 million. The report puts that down to higher prices and lower manufacturing costs. The unit shipped less memory and storage than it had a year before.

Bufetico puts every company on 30 exchanges through the same six layers, updated after each market close. The growth figures use median year-over-year change, and the shape of each series is tested separately. What passes is decided by thresholds you set yourself. See it.

← All articles